CN118984879A

CN118984879A - Methods for monitoring cancer using fragmentation patterns

Info

Publication number: CN118984879A
Application number: CN202380027237.6A
Authority: CN
Inventors: K·伦巴德; L·基弗; J·凯利; A·莱尔; N·C·德拉科波利; L·里纳尔迪
Original assignee: Delphi Diagnostics
Current assignee: Delphi Diagnostics
Priority date: 2022-03-17
Filing date: 2023-03-17
Publication date: 2024-11-19

Abstract

The present disclosure provides methods and systems for detecting, monitoring, diagnosing, predicting cancer status using analysis of a free DNA (cfDNA) fragmentation profile in a sample obtained from a patient, determining the likelihood of the presence of cancer and administering therapy to the patient, the methods and systems determining a ratio of short fragments to long fragments and a fragment size distribution from the fragmentation profile, and determining a divergence score based on the ratio of the short fragments to long fragments in the sample as related to the ratio of the sample from a healthy patient, and determining a monitoring score for the sample by a machine learning model based on the fragmentation score, the divergence score, and model weights, the monitoring score indicating the level of tumorigenic nucleic acid in cfDNA of the sample.

Description

Methods for monitoring cancer using fragmentation patterns

Cross Reference to Related Applications

The application claims the benefit of U.S. provisional patent application serial No. 63/320,906 filed on 3/17 of 2022, the entire contents of which are incorporated herein by reference in their entirety.

Technical Field

The present invention relates generally to genetic analysis, and more particularly to a method and system for analyzing free DNA (cfDNA) fragments to predict the fraction of tumor-derived DNA modules (ctDNA burden) and detect cancer in a subject.

Background

The onset and death of most human cancers worldwide are due to late diagnosis of these diseases, where treatment is less effective. In addition, after diagnosing cancer, predicting cancer progression and patient response to treatment is challenging, which can further impair the success rate of cancer treatment. Unfortunately, biomarkers that are clinically proven to be useful in widely diagnosing patients with cancer and predicting effective treatment of the patient are not widely available.

The fraction of tumor-derived DNA molecules in plasma (ctDNA burden) is a tool that can be used to describe the total tumor burden of patients with cancer. Previous work showed that ctDNA burden in individual patients is affected by a number of factors, including the tissue and stage of origin of the tumor and vascularization and perfusion. Thus, the ctDNA burden is higher in patients with more advanced cancers than in patients with earlier cancers. Similarly, patients with cancers that are located in tissue with high turnover rates and that directly enter the blood stream (e.g., colorectal cancer) typically have higher ctDNA loads than patients with less vascular, slower growing tumors. ctDNA burden may vary over time as tumors are exposed to treatment and die (decrease) and subsequently acquire a mechanism of resistance to treatment and grow (increase). However, previous studies lack the ability to effectively predict ctDNA burden and utilize the predicted ctDNA burden as a tool for diagnosing cancer, predicting disease progression and therapeutic response, and determining the overall survival of patients diagnosed with cancer.

Disclosure of Invention

The present disclosure provides methods and systems that utilize analysis of cfDNA to monitor cancer progression and predict total survival of a subject by scoring cfDNA fragmentation patterns obtained by analysis of cfDNA fragments in a sample obtained from the subject. The scoring method generates features that can be used to train a machine learning model to predict biomarkers that can be used to monitor cancer progression, evaluate patient responses to treatment, and predict overall viability of the subject.

Thus, in one embodiment, the invention provides a method of monitoring cancer. The method comprises the following steps:

determining a free DNA (cfDNA) fragmentation profile of a sample from the subject;

calculating a score based on the cfDNA fragmentation profile, the score indicating a likelihood of cancer being present in the subject;

Determining a ratio of short fragments to long fragments and a fragment size distribution according to the fragmentation profile;

calculating a divergence score based on the ratio of the short segments to the long segments;

Determining a set of model weights based on the segment size distribution;

Training a machine learning model using a set of features extracted from a plurality of fragmented atlases of a plurality of subjects; and

A monitoring score of the sample is determined by the machine learning model based on the monitoring score, the divergence score, and the model weight, the monitoring score indicating a level of tumorigenic nucleic acids in cfDNA of the sample.

In some aspects, the cfDNA fragmentation profile is determined by: cfDNA fragmentation profile was determined by: obtaining cfDNA fragments from the subject and isolating the cfDNA fragments; sequencing the cfDNA fragments to obtain sequenced fragments; mapping the sequenced fragments to a genome to obtain a window of mapped sequences; and analyzing a window of the mapped sequences to determine cfDNA fragment lengths and generate the cfDNA fragmentation profile.

In another embodiment, the invention provides a method of determining at least one of total survival, progression-free survival, or time to progression of a subject having cancer, comprising. The method comprises the following steps:

Determining a set of model weights based on the segment size distribution;

training a machine learning model using a set of features extracted from the fragmented atlases of the plurality of subjects;

Determining, by the machine learning model, a monitoring score for the sample based on the score, the divergence score, and the model weight, thereby indicating a likelihood of cancer progression in the subject, the monitoring score indicating a level of tumorigenic nucleic acid in cfDNA of the sample; and

Determining at least one of a likelihood of total survival of the subject, a progression-free survival, or a time until progression based on the monitoring score, thereby determining total survival of the subject. In yet another aspect, the invention provides a method of treating a subject having cancer.

In yet another embodiment, the invention provides a system for monitoring cancer in a subject. The system comprises:

A memory; and

One or more processors coupled to the memory, the one or more processors configured to perform operations that cause a computer system to:

Determining a set of model weights based on the segment size distribution;

Training a machine learning model using a set of features extracted from the fragmented atlases of the plurality of subjects; and

Determining, by the machine learning model, a monitoring score for the sample based on the score, the divergence score, and the model weight, thereby indicating a likelihood of cancer progression in the subject, the monitoring score indicating a level of tumorigenic nucleic acid in cfDNA of the sample.

In another embodiment, the invention provides a non-transitory computer readable storage medium encoded with a computer program. The computer program comprises instructions which, when executed by one or more processors, cause the one or more processors to perform operations for performing the method of the invention.

In yet another embodiment, the present invention provides a computing system. The system includes a memory and one or more processors coupled to the memory, wherein the one or more processors are configured to perform operations that implement the methods of the present invention.

In yet another embodiment, the present invention provides a system for genetic analysis and assessment of cancer, the system comprising: (a) A sequencer configured to generate a Whole Genome Sequencing (WGS) dataset of a sample; and (b) the non-transitory computer readable storage medium and/or computer system of the present invention.

Drawings

Fig. 1 is a block diagram illustrating a process for training a machine learning model to generate DELFI Monitoring Scores (DMS). The DMS can be used to monitor cancer in a patient diagnosed with cancer, determine total survival, determine progression-free survival, and determine time until progression. DMS can also be used to diagnose cancer in a patient and determine the cancer treatment administered to the patient.

Fig. 2 is a graphical diagram showing a comparison between an observed MAF trajectory and a trajectory determined based on DMS values.

Fig. 3 is a graphical diagram showing a comparison between MAF trajectories observed without reference to a point in time or a patient from which a given sample originated and trajectories determined based on DMS values.

Fig. 4 is a graphical diagram showing a comparison between an observed MAF trajectory and a trajectory determined based on DMS values. The patient group with the observed MAF trajectory in this figure suffers from a different type of cancer than the patient group used to train the model that generated the DMS values.

Figure 5 is a graphical diagram showing progression free survival data obtained by predicted DMS for two different groups with different types of cancer.

Fig. 6 is a graphical diagram showing the total survival obtained using the predicted DMS at the pre-treatment time point for a group of cancer patients. The graphical diagram also shows the total survival of the cancer patient groups undergoing incomplete and complete resections.

Fig. 7 is a graphical diagram showing progression free survival of a group of cancer patients who reached an optimal 0% of cloned variant MAF when assessed after the first treatment. The graph also shows a comparison between the predicted DMS and the MAF measured by ddPCR.

Fig. 8 is an example computer 800 that may be used to implement the training algorithm shown in fig. 1 and generate DMS values.

Fig. 9A-9B study design and patient flow chart. Fig. 9A is a CAIRO study design flow chart. After confirmation of qualification (including non-resectable status of liver metastasis defined by the central panel), KRAS (exons 2, 3 and 4), NRAS (exons 2 and 3) and BRAF mutation status were assessed in the tissue samples. Patients with RAS/BRAF mutant tumors were randomly assigned between either dual chemotherapy plus bevacizumab (panel 1) or triple chemotherapy plus bevacizumab (panel 2). Patients with RAS/BRAF wild-type tumors were randomly assigned between the dual chemotherapies plus bevacizumab (panel 3) or panitumumab (panel 4). In this transformation study, blood draws from patients in group 1 (mutant) and group 3 (wild type) were processed and analyzed. Fig. 9B is a patient flow chart. The number of patients and samples included in the study and the reasons for exclusion are depicted.

FIGS. 10A-10D illustrate that DELFI tumor scores (DELFI-TF) are a mutation agnostic method for metastatic disease monitoring. Fig. 10A shows that hotspot mutations in KRAS, NRAS and BRAF of tumors from untreated non-operable liver-only mCRC patients incorporated in the CAIRO phase III trial were tested. Blood samples were taken at baseline, during treatment and at the time of disease progression or last follow-up. Patients with driving mutations were monitored using ddPCR and DELFI-TF assays. Patients with wild-type KRAS, NRAS and BRAF were monitored only with DELFI-TF. Fig. 10B shows cfDNA isolation using plasma aliquots from patients with tissue-confirmed RAS/BRAF mutant mCRC. ddPCR and low coverage WGS were performed using duplicate cfDNA samples from each time point. WGS fragment sequencing statistics were calculated for each sample at a given time point. A Bayesian probability model (Bayesian probabilistic model) was trained on the invoked MAFs from ddPCR reads of tumor-specific RAS/BRAF variants in all longitudinal cfDNA samples to generate DELFI-TF values. FIG. 10C is a heat map of genomic features depicting the deviation in cfDNA fragment ratios and chromosome arm level z scores across baseline and in-treatment time points for 128 patients, as well as DELFI-TF values and clinical and demographic characteristics. Fig. 10D shows that cfDNA whole genome fragmentation patterns in 504 non-overlapping 5-Mb genomic regions at baseline and near the time point of the second imaging assessment by RECIST1.1 show significant heterogeneity in patients exhibiting disease progression at baseline and compared to patients undergoing stable disease or radiation response following initial first-line systemic therapy.

Figures 11A-11F show how DELFI-TF predicts tumor fraction in blood of patients with advanced disease who received systemic treatment. Figure 11A shows that patients with mCRC exhibit a range of DELFI-TF values at baseline that is wider than the range of DELFI-TF values for non-cancer controls. The upper 95% CI limit (grey dashed line) in the non-cancer control represents the DELFI-TF limit (0.006) for the blank. Fig. 11B shows that DELFI-TF correlated strongly with MAF measured by ddPCR at all study time points (n=692, pearson correlation (Pearson correlation), r=0.85, p=3.9 e-89). Plasma time points where MAF was undetectable (n=60, red) exhibited a broad range of DELFI-TF values. FIG. 11C shows that cfDNA fragmentation patterns (bottom) of patients with wild-type mCRC exhibit short-to-long ratio aberrations, even in the context of tumor copy neutral regions (top) that are binned across 100kb in matched tissue samples. Fig. 11D shows that in colorectal cancer, plasma tumor scores assessed by DELFI-TF (blue) and RAS/BRAF MAF (orange) values correlated with cfDNA copy number changes in genomic regions carrying frequently deleted (MBD 1) genes or amplified (PLGC 1) genes. Fig. 11E and 11F show that the relative coverage at the TSS location of the group of 890 genes highly expressed in colorectal cancer shows that the trough of the baseline sample (brown) is significantly deeper than the trough of the sample under treatment (purple) (Wilcoxon test, p < 0.001), indicating that on average, the tumor score is lower at the time point after initiation of treatment.

Figures 12A-12C show cfDNA tumor scores assessed by whole genome sequencing methods. Fig. 12A shows that patients with metastatic colorectal cancer exhibit a range of ichorCNA values at baseline that is wider than the range of ichorCNA values for the non-cancer control. The upper 95% confidence interval limit for the non-cancer control was 0.017 (grey dashed line). Fig. 12B shows that DELFI-TF correlates with tumor scores at time points where mutant allele fractions of RAS/BRAF assessed by ddPCR were undetectable in the mutant panel of studies measured by ichorCNA (spearman correlation (Spearman correlation), rho=0.54, p < 0.001). Fig. 12C shows the relative coverage of Transcription Start Site (TSS) positions at baseline (brown), during systemic treatment (purple), post-metastatic excision (red) and at disease recurrence (black) for patient 65 for the group consisting of genes highly expressed in colorectal cancer. The change in valley depth reflects the dynamic change in plasma tumor score at the longitudinal time point.

FIGS. 13A-13G show DELFI-TF is a non-invasive biomarker for metastatic disease burden and systemic treatment response. Fig. 13A shows that DELFI-TF (blue) and RAS/BRAF MAF (orange) values at baseline show moderate correlation with SLD of liver target lesions in pre-treatment imaging scans (DELFI-TF, spearman correlation, rho=0.49, p < 0.001; MAF, spearman correlation, rho=0.48, p < 0.001). Fig. 13B shows that DELFI-TF (blue) and RAS/BRAF MAF (orange) values at baseline show no significant correlation with baseline levels of carcinoembryonic antigen (DELFI-TF, spearman correlation, rho=0.10, p=0.427; MAF, spearman correlation, rho=0.15, p < 0.236). Fig. 13C shows that the DELFI-TF and MAF values at baseline are significantly lower in patients with a later confirmed Partial Response (PR) or Complete Response (CR) obtained by two consecutive RECIST 1.1 measurements (DELFI-TF, wilcoxon test, p=0.048; MAF, wilcoxon test, p=0.017). Figure 13D shows that DELFI-TF and MAF values at baseline were significantly lower in patients treated with either total ablation (orange) or partial ablation (green) surgery after systemic treatment (DELFI-TF, kruskal-Wallis test, p=0.037; MAF, kruskal-Wallis test, p=0.011). Figure 13E shows that colorectal cancer patients with allogeneic metastases (grey) show lower tumor scores at baseline as assessed by DELFI-TF than patients presenting with concurrent metastases (green) (wilcoxon test, p < 0.001). Fig. 13F shows that in the portrait imaging scan, liver metastasis is highlighted by blue circles (top). cfDNA tumor score dynamics (DELFI-TF, MAF) and SLD values (bottom) for study patient 11 are shown. Treatment is indicated by shaded bars. The purple dotted line indicates the time of primary tumor resection within weeks after liver metastasis removal. DELFI-TF and ddPCR MAF directed against KRAS G12D mutations accurately track disease burden dynamics before and after complete excision. Figure 13G shows that patients eventually experiencing disease progression (top) more often show DELFI-TF increase in value at the longitudinal time point than patients who never exhibited disease progression (bottom).

Figures 14A-14D show DELFI-TF and colorectal cancer clinical properties. Fig. 14A shows that the tumor scores assessed by DELFI-TF and MAF at baseline were equal between colon cancers on the left (brown) or right (beige) side (DELFI-TF, wilcoxon test, p=0.329; MAF, wilcoxon test, p=0.515). Fig. 14B shows that the tumor scores assessed by DELFI-TF at baseline are equal between colon cancers with driving RAS/BRAF mutations (red) or wild type (pink) (wilcoxon test, p=1). Fig. 14C shows that the tumor score at baseline assessed by DELFI-TF and MAF was higher in patients eventually undergoing disease progression (green) than in patients not undergoing disease progression (red) (DELFI-TF, wilcoxon test, p=0.03; MAF, wilcoxon test, p=0.02). The sum of maximum diameters at baseline was not different between the once progressors (green) and the never progressors (red) (wilcoxon test, p=0.94). Fig. 14D shows a waterfall plot depicting an objective clinical response through a Sum of Longest Diameters (SLD) change. Red, DELFI-TF slope above median. Cyan, DELFI-TF slope below median.

Figures 15A-15B show DELFI-TF dynamics of study patients. Fig. 15A shows the RAS/BRAF wild type panel. Fig. 15b, ras/BRAF mutant panel.

Figure 16 shows that DELFI-TF dynamics correlate with longitudinal clinical outcome. DELFI-TF slope was calculated using all time points available to patients who had one blood draw at least at baseline and within 60 days of disease progression (n=81). On the left, DELFI-TF slope is colored based on the results below (cyan) or above (red) median DELFI-TF slope. On the right, swim diagrams, which cover RECIST 1.1, liquid biopsies, surgical and death events for patients ordered in weeks according to time in study. Each bar represents the interval between study enrollment and death or last follow-up. The strip section was colored according to RECIST 1.1 readings.

Figures 17A-17D show that baseline DELFI-TF and DELFI-TF slopes correlate with Progression Free Survival (PFS) and total survival (OS). Fig. 17A shows Kaplan-Meier curves (n=128) of PFS plotted against baseline DELFI-TF values below (orange) or above (blue) lower quartile among patients with RAS/BRAF mutants and wild-type metastatic colorectal cancer (log rank p=0.015). Figure 17B shows Kaplan-Meier curves (n=81) of PFS plotted against DELFI-TF slopes below (orange) or above (blue) median (log rank p < 0.001) among patients with at least one blood draw within 60 days of disease progression. Figure 17C shows Kaplan-Meier curves (n=42) of PFS plotted against DELFI-TF slope below (orange) or above (blue) median (log rank p < 0.001) among patients experiencing imaging response or stable disease for longer than 12 months. Fig. 17D shows Kaplan-Meier curves (n=81) of OS plotted against DELFI-TF slopes below (orange) or above (blue) median (log rank p < 0.001) among patients with at least one blood draw within 60 days of disease progression.

Figures 18A-18D show imaging and plasma biomarkers of survival outcome for patients with metastatic colorectal cancer. Fig. 18A shows that Kaplan-Meier curves of Progression Free Survival (PFS) obtained from imaging responses assessed by RECIST 1.1 show no survival differences between patients with Partial Response (PR) (orange) or Stable Disease (SD) (blue). PD, progressive disease. Fig. 18B shows Kaplan-Meier curves (n=65) for PFS plotted against baseline MAF below (orange) or above (blue) lower quartile values for RAS/BRAF among patients with RAS/BRAF mutant metastatic colorectal cancer (log rank p=0.003). Fig. 18C shows Kaplan-Meier curves (n=127) (log rank p=0.067) of PFS plotted against baseline carcinoembryonic antigen values below (orange) or above (blue) lower quartile values among patients with metastatic colorectal cancer. Fig. 18D shows Kaplan-Meier curves (n=81) of total survival (OS) plotted against surgical status and DELFI-TF slope among patients with at least one blood draw within 60 days of disease progression (log rank p < 0.001).

Detailed Description

Described herein is a non-invasive method for monitoring cancer in a subject having cancer and predicting overall survival, progression-free survival, and time until progression. cfDNA in the blood may provide a non-invasive way to monitor disease in patients with cancer. As demonstrated herein, DNA evaluation of early intercept fragments (DELFI) was used to evaluate the whole genome fragmentation pattern of cfDNA of patients with various types of cancer as well as healthy individuals. Assessment of cfDNA included scoring methods. The defined score (also referred to herein as the "DELFI monitor score") is determined based on cfDNA fragmentation patterns obtained using cfDNA fragments of a given patient sample. Assessing cfDNA using the methods described herein may also provide a method for monitoring cancer, which may increase the chance of successful treatment and improve the outcome of patients with cancer.

Before describing the compositions and methods of the present invention, it is to be understood that this invention is not limited to the particular methods and systems described as such methods and systems may vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present invention will be limited only by the appended claims.

As used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "the method" includes one or more methods and/or steps of the type described herein that will become apparent to those skilled in the art upon reading the present disclosure and the like.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, the preferred methods and materials are now described.

The present disclosure provides innovative methods and systems for analyzing cfDNA to monitor, detect, or otherwise evaluate cancer. As indicated by the previous study, on average, cancer-free individuals had longer cfDNA fragments (average size 167.09 bp) while individuals with cancer had shorter cfDNA fragments (average size 164.88 bp). The methods described herein allow for simultaneous analysis of large numbers of abnormalities in cfDNA by whole genome analysis of cfDNA fragmentation patterns.

Thus, in one embodiment, the invention provides a method of monitoring cancer in a subject. The method comprises the following steps:

Determining a set of model weights based on the segment size distribution;

In an embodiment, the invention provides a method of treating a subject having cancer. The method comprises the following steps:

a) Detecting cancer in the subject using the methods of the invention, or determining the total survival of the subject using the methods of the invention; and

B) Administering a cancer treatment to the subject, thereby treating the subject.

In another embodiment, the invention provides a method of monitoring cancer in a subject. The method comprises the following steps:

a) Detecting cancer in the subject using the methods of the invention, or determining the total survival of the subject using the methods of the invention;

b) Administering a cancer treatment to the subject; and

C) The method of the invention is used to determine the overall survival of the subject after administration of the cancer treatment, thereby monitoring the subject for cancer.

The methods described herein utilize cfDNA fragmentation patterns. As used herein, the term "fragmentation profile" (in some aspects, determining cfDNA fragmentation profile of a mammal) may be used to identify a mammal as having cancer. For example, cfDNA fragments obtained from a mammal (e.g., from a sample obtained from a mammal) may be subjected to low coverage whole genome sequencing, and the sequenced fragments may be mapped to the genome (e.g., in non-overlapping windows) and evaluated to determine a cfDNA fragmentation profile. The cfDNA fragmentation profile of a mammal having cancer is more heterogeneous (e.g., in fragment length) than the cfDNA fragmentation profile of a healthy mammal (e.g., a mammal not having cancer).

The cfDNA fragmentation profile may comprise one or more cfDNA fragmentation patterns. The cfDNA fragmentation pattern may comprise any suitable cfDNA fragmentation pattern. Examples of cfDNA fragmentation patterns include, but are not limited to, fragment size density, median fragment size, fragment size distribution, ratio of small cfDNA fragments to large cfDNA fragments, and coverage of cfDNA fragments. In some aspects, the cfDNA fragmentation profile may be a whole genome cfDNA profile (e.g., a whole genome cfDNA profile in a window throughout the genome). In some aspects, the cfDNA fragmentation profile may be a targeting region profile. The targeting region may be any suitable portion of the genome (e.g., a chromosomal region). Examples of chromosomal regions for which cfDNA fragmentation patterns can be determined as described herein include, but are not limited to, a portion of a chromosome (e.g., a portion of 2q, 4p, 5p, 6q, 7p, 8q, 9q, 10q, 11q, 12q, and/or 14 q) and a chromosomal arm (e.g., a chromosomal arm of 8q, 13q, 11q, and/or 3 p). In some cases, the cfDNA fragmentation profile may comprise two or more targeting region profiles.

In various aspects, cfDNA obtained from a sample is isolated and fragments of a particular size range are utilized in an assay. In some aspects, the analysis excludes fragment sizes of less than about 10bp, 50bp, 100bp, or 105bp and greater than about 220bp, 250bp, 300bp, 350bp, or more. In some aspects, the analysis excludes fragment sizes less than 105bp and greater than 170 bp. In some aspects, the analysis excludes fragment sizes of less than about 230bp, 240bp, 250bp, 260bp, and greater than about 420bp, 430bp, 440bp, 450bp, or more. In some aspects, the analysis excludes fragment sizes less than 260bp and greater than 440 bp.

In some aspects, cfDNA fragmentation patterns can be determined by: processing a sample from a subject comprising cfDNA fragments into a sequencing library; subjecting the sequencing library to low coverage whole genome sequencing to obtain sequenced fragments; mapping the sequenced fragments to a genome to obtain a window of mapped sequences; and analyzing a window of the mapped sequences to determine cfDNA fragment lengths.

In some aspects, cfDNA fragmentation patterns can be determined by: obtaining cfDNA fragments from a subject and isolating the cfDNA fragments; sequencing the cfDNA fragments to obtain sequenced fragments; mapping the sequenced fragments to a genome to obtain a window of mapped sequences; and analyzing a window of the mapped sequences to determine cfDNA fragment lengths and generate the cfDNA fragmentation profile.

The method of the invention is based on low coverage whole genome sequencing and analysis of isolated cfDNA. In one aspect, the data used to develop the methods of the invention are based on shallow whole genome sequence data (1-2 x coverage).

In some aspects, the mapped sequences are analyzed in a non-overlapping window covering the genome. Conceptually, the size of the window may range from thousands to millions of bases, producing hundreds to thousands of windows in the genome. A 5Mb window was used to evaluate cfDNA fragmentation patterns, as these windows would provide over 20,000 reads per window, even within a limited amount of 1-2x genome coverage. Within each window, the coverage and size distribution of cfDNA fragments was checked. In some aspects, the whole genome pattern from the individual can be compared to the whole genome pattern of a reference population to determine whether the pattern is likely healthy or cancer-derived.

In certain aspects, the mapped sequences comprise tens to thousands of genome windows, such as 10, 50, 100 to 1,000, 5,000, 10,000, or more windows. Such windows may be non-overlapping or overlapping, and may contain about 100, 200, 300, 400, 500, 600, 700, 800, 900 or 1000 base pairs.

In various aspects, cfDNA fragmentation patterns are determined within each window. Accordingly, the present invention provides methods for determining a cfDNA fragmentation profile of a subject (e.g., in a sample obtained from a subject).

In some aspects, cfDNA fragmentation patterns can be used to identify changes (e.g., alterations) in cfDNA fragment length. The alteration may be a whole genome alteration or an alteration of one or more targeted regions/loci. The target region may be any region containing one or more cancer specific changes. In some aspects, cfDNA fragmentation patterns can be used to identify (e.g., simultaneously identify) about 10 to about 500 changes (e.g., about 25 to about 500, about 50 to about 500, about 100 to about 500, about 200 to about 500, about 300 to about 500, about 10 to about 400, about 10 to about 300, about 10 to about 200, about 10 to about 100, about 10 to about 50, about 20 to about 400, about 30 to about 300, about 40 to about 200, about 50 to about 100, about 20 to about 100, about 25 to about 75, about 50 to about 250, or about 100 to about 200 changes).

In various aspects, the cfDNA fragmentation profile may comprise a cfDNA fragment size pattern. The cfDNA fragments may be of any suitable size. For example, in some aspects, the cfDNA fragment may be about 50 base pairs (bp) to about 400bp in length. As described herein, a subject with cancer may have a cfDNA fragment size pattern that contains median cfDNA fragment sizes that are shorter than the median cfDNA fragment sizes of healthy subjects. Healthy subjects (e.g., subjects not suffering from cancer) may have cfDNA fragment sizes with median cfDNA fragment sizes of about 166.6bp to about 167.2bp (e.g., about 166.9 bp). In some aspects, a subject with cancer may have a cfDNA fragment size that is about 1.28bp to about 2.49bp (e.g., about 1.88 bp) shorter than the cfDNA fragment size of a healthy subject on average. For example, a subject with cancer may have a cfDNA fragment size with a median cfDNA fragment size of about 164.11bp to about 165.92bp (e.g., about 165.02 bp).

In some aspects, the dinuclear cfDNA fragments may be about 230 base pairs (bp) to about 450bp in length. As described herein, a subject with cancer may have a pattern of dinuclear corpuscle cfDNA fragment sizes that contains a median dinuclear corpuscle cfDNA fragment size that is shorter than the median dinuclear corpuscle cfDNA fragment size of a healthy subject. In some aspects, on average, cancer-free subjects have longer cfDNA fragments (average size 334.75 bp) within the dinuclear corpuscle range, while subjects with cancer have shorter dinuclear corpuscle cfDNA fragments (average size 329.6 bp). As such, a healthy subject (e.g., a subject not suffering from cancer) may have a dinuclear, small cfDNA fragment size with a median cfDNA fragment size of about 334.75 bp. In some aspects, a subject with cancer may have a dinuclear, small cfDNA fragment size that is shorter than the dinuclear, small cfDNA fragment size of a healthy subject. For example, a subject with cancer may have a dinuclear, small cfDNA fragment size of about 329.6bp in median cfDNA fragment size.

The cfDNA fragmentation profile may comprise a cfDNA fragment size distribution. As described herein, a subject with cancer may have a cfDNA size distribution that is more variable than that of a healthy subject. In some aspects, the size distribution may be within the targeted region. Healthy subjects (e.g., subjects not suffering from cancer) can have a targeted region cfDNA fragment size distribution of about 1 or less than about 1. In some aspects, a subject with cancer may have a targeting region cfDNA fragment size distribution that is longer (e.g., 10bp, 15bp, 20bp, 25bp, 30bp, 35bp, 40bp, 45bp, 50bp, or more, or any number of base pairs in between) than the targeting region cfDNA fragment size distribution of a healthy subject. In some aspects, a subject with cancer may have a targeting region cfDNA fragment size distribution that is shorter (e.g., 10bp, 15bp, 20bp, 25bp, 30bp, 35bp, 40bp, 45bp, 50bp, or more, or any number of base pairs in between) than the targeting region cfDNA fragment size distribution of a healthy subject. In some aspects, a subject with cancer may have a target region cfDNA fragment size distribution that is about 47bp to about 30bp smaller than the target region cfDNA fragment size distribution of a healthy subject. In some aspects, a subject with cancer may have a targeting region cfDNA fragment size distribution with cfDNA fragments that differ in average by 10bp, 11bp, 12bp, 13bp, 14bp, 15bp, 17bp, 18bp, 19bp, 20bp, or more in length. For example, a subject with cancer may have a distribution of target region cfDNA fragment sizes where the cfDNA fragments average differ in length by about 13 bp. In some aspects, the size distribution may be a whole genome size distribution.

The cfDNA fragmentation profile may comprise a ratio of small cfDNA fragments to large cfDNA fragments and a correlation of fragment ratios to reference fragment ratios. As used herein, with respect to the ratio of small cfDNA fragments to large cfDNA fragments, the small cfDNA fragments may be about 100bp in length to about 150bp in length. As used herein, with respect to the ratio of small cfDNA fragments to large cfDNA fragments, the large cfDNA fragments may be about 151bp in length to 220bp in length. As described herein, the fragment ratio correlation (e.g., correlation of cfDNA fragment ratio to reference DNA fragment ratio (e.g., DNA fragment ratio from one or more healthy subjects)) of a subject with cancer may be lower than the fragment ratio correlation of a healthy subject (e.g., 1/2, 1/3, 1/4, 1/5, 1/6, 1/7, 1/8, 1/9, 1/10, or less of the fragment ratio correlation of the healthy subject). The fragment ratio correlation (e.g., correlation of cfDNA fragment ratio to reference DNA fragment ratio (e.g., DNA fragment ratio from one or more healthy subjects)) for a healthy subject (e.g., a subject not suffering from cancer) can be about 1 (e.g., about 0.96). In some aspects, the fragment ratio correlation (e.g., correlation of cfDNA fragment ratio to reference DNA fragment ratio (e.g., DNA fragment ratio from one or more healthy subjects)) of a subject with cancer may be about 0.19 to about 0.30 (e.g., about 0.25) on average lower than the fragment ratio correlation (e.g., correlation of cfDNA fragment ratio to reference DNA fragment ratio (e.g., DNA fragment ratio from one or more healthy subjects)) of a healthy subject.

The method of the invention further comprises predicting Mutant Allele Fractions (MAF) based on the cfDNA fragmentation profile. Mutated MAFs in DNA are common values reported by diagnostic tests against tumors and represent the fraction of DNA molecules analyzed that contain the mutation of interest. For tumor-derived variants identified in circulating free DNA (cfDNA), MAF represents the fraction of all cfDNA containing the variant. cfDNA is a combination of tumor-derived DNA and normal cell-derived DNA, and therefore, the MAF value of the clonal somatic variant is captured as a fraction of the tumor-derived cfDNA. This MAF is associated with and can therefore be used as a surrogate for the circulating tumor DNA fraction.

In various embodiments, the present invention can use the predicted MAF to detect cancer in a subject, predict disease prognosis, predict response to treatment, and/or assess overall survival of a subject. FIG. 1 is a block diagram illustrating a process for training a machine learning model to predict MAF. The predicted MAF may be referred to as DELFI Monitoring Score (DMS). At 102, a DELFI score is calculated based on the cfDNA fragmentation profile. The DELFI score indicates that the fragmentation pattern is similar to a prototype individual with cancer or an individual without cancer. In some aspects, calculating DELFI the score comprises: i) Determining the ratio of short cfDNA fragments to long cfDNA fragments of the sample, ii) determining the Z-score of cfDNA fragments of the sample by chromosome arm, iii) quantifying cfDNA fragment density using computational mixture model analysis, and iv) processing the output of i) -iii) using machine learning model to define the score. In various aspects, the score is used to determine the likelihood of overall survival of the subject.

In one illustrative example (example 1), in a multi-cancer cohort, the inventors calculated a mixed model of short-to-long segment ratios calculated in 5MB bins, Z-scores calculated in chromosome arms, and cfDNA segment sizes for each individual according to low coverage whole genome sequencing. Using these features as inputs, the inventors fit cross-validated gradient-enhanced machines to each person's cancer status (cancer/non-cancer). The output of this model is a score ranging from 0 to 1, with a high value indicating a stronger signal for cancer and a low value indicating a more similarity to non-cancer. Scores generated using these techniques may use features to train a machine learning model to generate a DMS.

At 104, DELFI divergences may be calculated. In some aspects, DELFI divergence may be equal to 1 minus the correlation between the binned and average centered short-to-length ratio for a given sample and the binned and average centered short-to-length ratio for a healthy sample. For example, a healthy sample may be equal to the median of the short-to-long ratios of the bins of the reference group containing only healthy samples and centered on the average. As used herein, the average centered short-to-long ratio is the binned short-to-long ratio minus the total average.

At 106, a set of weights to calculate a hybrid model may be determined. In some aspects, the hybrid model may be a vector containing 11 weights that summarize the fragmentation distribution in the sample. Weights from the mixture model are estimated using bayesian mixture normal distributions of empirical fragment size distributions.

At 108, a regression model may be trained on the measured MAFs of the individuals such that the model learns the characteristics of the DELFI score, DELFI divergence, and mixed model weights of the sample that contribute to the known MAFs of the sample. For example, the MAF may be related to tumor burden, e.g., estimated by MAF. In some aspects, the regression model may be a bayesian hierarchical regression model that includes a plurality of layers, with each layer including more predictors. At run-time, the model takes DELFI scores, DELFI divergences, and mixed model weights as inputs, and outputs the predicted MAF. Training is accomplished by "leave one patient" cross-validation. In such a cross-validation scheme, data for each patient is set aside in turn, the model is trained on the remaining samples, and the trained model is then used to generate predictions for the set aside samples. In one example of a model, MAF is a random variable of β distribution, and the model assumes that the expected MAF for a given sample is functionally related to the described feature by the inverse of the feature matrix multiplied by the vector of regression coefficients plus a patient-specific random intercept that accounts for intra-patient correlation between measurements.

At 110, the trained model may be validated to confirm that it reaches a desired level of accuracy. In various embodiments, the trained models may be evaluated statistically and clinically. For example, the quality of the generated predictions may be evaluated by evaluating the correlation of the predicted tumor burden with the observed tumor burden value. Other examples of verification protocols performed to evaluate a trained model include observing a longitudinal graph showing measured tumor burden values and superimposed predicted tumor burden values and evaluating the relationship between a patient's tumor burden prediction and time to death.

The model may also be validated in a clinical setting to understand the predicted clinical utility. To clinically verify the model, two untreated metastatic groups were obtained. The cohort began anti-tumor treatment with chemotherapy or targeting agent, and it was determined that the predicted ctDNA burden (represented by MAF) could predict the survival of each patient in the cohort at baseline and when blood was drawn after the first treatment. Observed MAF data were obtained for 76 patients with metastatic colorectal cancer (mCRC) and 17 patients with metastatic non-small cell lung cancer (mNSCLC). All patient samples were analyzed independently using DELFI monitor scoring method. For mCRC patients and mNSCLC patients, blood withdrawal after the first treatment was performed at 4-12 weeks and 1-3 weeks after treatment, respectively. MAF of the clonal variants was measured by digital droplet PCR (RAS/RAF variants) and depth-targeted NGS sequencing (EGFR) on mCRC patients and mNSCLC patients, respectively.

A Kaplan-Meier estimator was used to evaluate the predicted value of a single threshold for the modeled ctDNA load. In some aspects, the threshold of the DMS may be selected for each group by leaving one patient cross-validated. In this analysis, one sample is removed and a threshold is selected that minimizes the log rank p-value. This process was repeated for each patient in the cohort and the median of all optimized thresholds from the cross-validation was selected as the final threshold for Kaplan-Meier estimation. The Cox proportional hazards model was also used, where available, to evaluate predictions of ctDNA load for continuous modeling of progression-free and total survival. In another aspect, other methods for determining the threshold may be used, such as using a reference set of individuals with no or low tumor scores.

Model verification results

FIG. 2 is a graphical diagram showing a comparison between observed longitudinal MAF data and predicted tumor burden generated by a model. In particular, the graph shows the relative MAF trajectory of mCRC patients obtained from observed MAF data and predicted MAF data. Each panel plots the observed longitudinal MAF data with black lines and the superimposed predictions from the model with blue lines. The figure also shows the boundaries of a 95% bayesian belief interval superimposed as a light blue shaded area on the black and blue lines. Each panel represents longitudinal data from one subject, and in most cases, the observed trajectory of the MAF for a given patient closely matches the predicted MAF (i.e., DMS).

Fig. 3 is a graphical diagram showing a comparison between the observed MAF and the predicted MAF without reference to a specific point in time or from which patient a given sample originated. Thus, the map contains the DMS of the entire mCRC group and the observed MAF. The solid diagonal lines contained in the figure represent absolute fairness lines between DMS and observed MAF. As shown, in most cases DELFI the monitored score matched quite well with the observed MAF. In a few cases where predictions are far from absolute fairness lines, there is evidence that the problem is not in the model, but in the measurement process, especially in samples where the measured MAF is low (i.e. less than 1%). For example, variants of interest may be more difficult to evaluate due to tumor and/or metastasis heterogeneity and clonal evolution that may occur at the time of treatment.

Fig. 4 is a graphical diagram showing a comparison between the observed MAF trajectory of mNSCLC patients and DMS values from the model training the mCRC cohort. Each panel plots the observed longitudinal MAF data with black lines and the superimposed predictions from the model with blue lines. The figure also shows the boundaries of a 95% bayesian belief interval superimposed as a light blue shaded area on the black and blue lines. Each panel represents longitudinal data from one subject, and in most cases, the observed trajectory of the MAF for a given patient closely matches the predicted MAF (i.e., DMS).

Comparison results indicate that models that train MAF data from patients with one type of cancer (e.g., mCRC patients) can be successfully applied to patients with a different type of cancer (e.g., mNSCLC groups). External applicability is a desirable feature of predictive models because the quality of predictions is generally high despite substantial differences between the two groups (cancer type, sequencing depth, etc.). External applicability of the predictive models described herein improves the efficiency of predictive model development and training by enabling one predictive model trained on a particular data set to be used to generate useful predictions for patients other than those contained in the training data set.

Fig. 5 is a graphical diagram showing in panel a the progression-free survival of a metastatic colorectal cancer group (i.e., mCRC group) and in panel B the progression-free survival of a metastatic lung cancer group (i.e., mNSCLC group). The figure shows the results from a Kaplan-Meier estimation that uses group-specific cross-validated thresholds to distinguish patients with high DMS and low DMS. In both groups, patients with DMS below the cross-validated threshold (DELFI monitor score (-)) at the first time point after treatment showed a longer progression-free survival than patients with DMS higher. These results indicate that DMS can distinguish a subset of patients that are more likely to survive without progression at the time point after the first treatment.

Fig. 6 is a graph showing the total lifetime of the DMS-based obtained mCRC group in panel a and the total lifetime of the DMS-based obtained post-treatment mCRC in panel B. To determine the ability of DMS to predict response to treatment, total lifetime data for the mCRC group was obtained. The life cycle data for each patient is marked with an indication of whether each patient undergoing surgery has undergone complete or incomplete resection. DMS at a time point prior to initiation of treatment was evaluated against separate cross-validated discrimination thresholds. As shown in panel a, the total lifetime of the patient with DMS below the threshold is longer than the total lifetime of the patient with DMS above the threshold. In addition, the DMS can further distinguish between patients in panel B who underwent incomplete resection or complete resection and patients with complete resection who have a longer predicted total survival. Taken together, the results demonstrate the feasibility of DMS to predict disease prognosis, even before treatment begins.

MAF of the cloned variants correlated with ctDNA burden and thus can be used as a quantitative indicator for patients to estimate fraction of tumor-derived plasma DNA and total tumor burden. However, during treatment, the genetic profile of the tumor may change under the selective pressure of the treatment. Thus, measuring MAF of only one variant is limited for measuring patient responses longitudinally. To evaluate the sensitivity of DMS to changes in tumor DNA during treatment, it was determined whether the MAF of the clonal variants in treatment measured as 0% of patients at the time point after the first treatment would benefit from analysis using DMS.

Fig. 7 is a graph showing in panel a the progression free survival of patients whose clonal variant MAF in the mCRC group was measured as 0% in the first post-treatment assessment and in panel B the DMS obtained by ddPCR relative to MAF of patients whose clonal variant MAF in the mCRC group was measured as 0% in the first post-treatment assessment. As shown in panel a, 43 patients with 0% maf in mCRC cohort were further isolated by DMS relative to progression free survival. Some of these patients were measured between 5% -15% dms as shown in panel B. This data suggests that predicted ctDNA loads based on whole genome evaluation of fragments may be more sensitive than conventional MAFs.

In addition, cox proportional risk analysis was performed on mCRC and mNSCLC groups to evaluate the predictions for continuous DMS. At both pre-treatment and first post-treatment time points, DMS can predict the total survival of the mCRC group (HR: 19.2, 95% ci:2.7-138.5, and HR:400.4, 95% ci:11.8-13581.0, respectively) and the progression-free survival in the mNSCLC group (HR: 67.3, 95% ci:1.1-4073.6, and HR:246.5, 95% ci:2.2-28030.9, respectively).

Example hardware implementation

Fig. 8 illustrates an example computer 800 that may be used to implement the training algorithm shown in fig. 1 and generate DMS values. For example, the computer 800 may contain a machine learning system that trains a machine learning model to generate DMS values as described above, or a portion or combination thereof in some embodiments. The computer 800 may be any electronic device running a software application derived from compiled instructions, including but not limited to a personal computer, server, smart phone, media player, electronic tablet, gaming machine, email device, and the like. In some embodiments, computer 800 may include one or more processors 802, one or more input devices 804, one or more display devices 806, one or more network interfaces 808, and one or more computer-readable media 812. Each of these components may be coupled by a bus 810, and in some embodiments, these components may be distributed across multiple physical locations and coupled by a network.

The display device 806 may be any known display technology including, but not limited to, a display device using Liquid Crystal Display (LCD) or Light Emitting Diode (LED) technology. The processor 802 may use any known processor technology, including but not limited to graphics processors and multi-core processors. The input device 804 may be any known input device technology including, but not limited to, a keyboard (including a virtual keyboard), a mouse, a trackball, a camera, and a touch sensitive pad or display. Bus 810 may be any known internal or external bus technology including, but not limited to ISA, EISA, PCI, PCI Express, USB, serial ATA, or firewire. Computer-readable media 812 may be any non-transitory media that participates in providing instructions to processor 804 for execution, including but not limited to non-volatile storage media (e.g., optical disks, magnetic disks, flash drives, etc.) or volatile media (e.g., SDRAM, ROM, etc.).

Computer-readable media 812 may contain instructions for implementing an operating system (e.g., macLinux). The operating system may be multi-user, multi-processing, multi-tasking, multi-threading, a real-time operating system, and the like. The operating system may perform basic tasks including, but not limited to: identifying an input from the input device 804; sending the output to a display device 806; keeping track of files and directories on computer-readable medium 812; controlling peripheral devices (e.g., disk drives, printers, etc.) that may be controlled directly or through an I/O controller; and manages traffic on bus 810. The network communication instructions 816 may establish and maintain a network connection (e.g., software for implementing a communication protocol such as TCP/IP, HTTP, ethernet, telephone, etc.).

The machine learning instructions 818 may include instructions that enable the computer 800 to act as a machine learning system and/or train a machine learning model to generate DMS values as described herein. Application 820 may be an application that uses or implements the processes described herein and/or other processes. The process may also be implemented in the operating system 814. For example, the application 820 and/or the operating system may create tasks in the applications described herein.

The described features may be implemented in one or more computer programs that are executable on a programmable system including at least one programmable processor coupled to receive data and instructions from, and to transmit data and instructions to, a data storage system, at least one input device, and at least one output device. A computer program is a set of instructions that can be used, directly or indirectly, in a computer to perform certain activities or bring about a certain result. A computer program can be written in any form of programming language, including compiled or interpreted languages (e.g., objective-C, java), and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

Processors suitable for the execution of a program of instructions may include, by way of example, both general and special purpose microprocessors, and the sole processor or one of multiple processors or cores of any kind of computer. Generally, a processor may receive instructions and data from a read-only memory or a random access memory or both. The essential elements of a computer may include a processor for executing instructions and one or more memories for storing instructions and data. In general, a computer may also contain, or be operatively coupled to communicate with, one or more mass storage devices for storing data files; such devices include magnetic disks, such as internal hard disks and removable disks; magneto-optical disk; an optical disc. Storage devices suitable for tangibly embodying computer program instructions and data may include all forms of non-volatile memory, including by way of example semiconductor memory devices (e.g., EPROM, EEPROM) and flash memory devices; magnetic disks, such as internal hard disks or removable disks; magneto-optical disk; CD-ROM and DVD-ROM discs. The processor and the memory may be supplemented by, or incorporated in, ASICs (application-specific integrated circuits).

To provide for interaction with a user, the features can be implemented on a computer having a display device (e.g., an LED or LCD monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer.

The features may be implemented on a computer system that includes a back-end component, such as a data server, or that includes a middleware component, such as an application server or an Internet server, or that includes a front-end component, such as a user computer having a graphical user interface or an Internet browser, or any combination of them. The components of the system can be connected by any form of digital data communication (e.g., a communications network) or medium of digital data communication. Examples of communication networks include, for example, telephone networks, LANs, WANs, and computers, and networks forming the internet.

The computer system may include a client and a server. The client and server may be generally remote from each other and may typically interact through a network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

One or more features or steps of the disclosed embodiments can be implemented using an API. An API may define one or more parameters that are passed between calling an application and other software code (e.g., operating system, library routines, functions) that provide services, provide data, or perform operations or computations.

An API may be implemented as one or more calls in program code that send or receive one or more parameters through a list of parameters or other structure based on calling conventions defined in an API specification document. The parameter may be a constant, a key, a data structure, an object class, a variable, a data type, a pointer, an array, a list, or another call. The API calls and parameters may be implemented in any programming language. The programming language may define a vocabulary and call conventions that a programmer will use to access functions that support the API.

In some embodiments, the API call may report to the application the capabilities of the device running the application, such as input capabilities, output capabilities, processing capabilities, power capabilities, communication capabilities, and the like.

While various embodiments have been described above, it should be understood that they have been presented by way of example, and not limitation. It will be apparent to persons skilled in the relevant art that various changes in form and detail can be made therein without departing from the spirit and scope. Indeed, it will be apparent to those skilled in the relevant art how to implement alternative embodiments after reading the foregoing description. For example, other steps may be provided, or steps may be eliminated, from the described flows, and other components may be added to, or removed from, the described systems. Accordingly, other embodiments are within the scope of the following claims.

Additionally, it should be understood that any figures highlighting functions and advantages are presented for exemplary purposes only. The disclosed methods and systems are each flexible and configurable enough that they can be utilized in a manner other than that shown.

Although the term "at least one" may be used frequently in the specification, claims and drawings, the terms "a", "an", "the" and the like also mean "at least one" or "the at least one" in the specification, claims and drawings.

Finally, the applicant intends that, according to 35 u.s.c.112 (f), only the claims containing the explicit language "means for..once again" or "steps for..once again" are to be interpreted. According to 35 u.s.c.112 (f), the claims that do not explicitly include the phrase "means for" or "steps for" will not be interpreted.

The presently described methods and systems may be used to detect, predict, treat, and/or monitor a cancer state in a subject. Any suitable subject (e.g., mammal) may be evaluated, monitored, and/or treated as described herein. Examples of some mammals that may be evaluated, monitored, and/or treated as described herein include, but are not limited to, humans, primates, such as monkeys, dogs, cats, horses, cows, pigs, sheep, mice, and rats. For example, a person having or suspected of having cancer may be assessed using the methods described herein, and optionally may be treated with one or more cancer treatments as described herein.

A subject suffering from or suspected of suffering from any suitable type of cancer may be monitored, assessed, and/or treated using the methods and systems described herein (e.g., by administering one or more cancer treatments to the subject). The cancer may be any stage of cancer. In some aspects, the cancer may be an early stage cancer. In some aspects, the cancer may be asymptomatic cancer. In some aspects, the cancer may be residual disease and/or recurrence (e.g., after surgical resection and/or after cancer therapy). The cancer may be any type of cancer. Examples of types of cancers that may be assessed, monitored and/or treated as described herein include, but are not limited to, lung cancer, colorectal cancer, prostate cancer, breast cancer, pancreatic cancer, cholangiocarcinoma, liver cancer, CNS cancer, gastric cancer, esophageal cancer, gastrointestinal stromal tumor (GIST), uterine cancer and ovarian cancer. Additional types of cancers include, but are not limited to, myeloma, multiple myeloma, B-cell lymphoma, follicular lymphoma, lymphocytic leukemia, and myelogenous leukemia. In some aspects, the cancer is a solid tumor. In some aspects, the cancer is a sarcoma, carcinoma, or lymphoma. In some aspects, the cancer is lung cancer, colorectal cancer, prostate cancer, breast cancer, pancreatic cancer, cholangiocarcinoma, liver cancer, CNS cancer, gastric cancer, esophageal cancer, gastrointestinal stromal tumor (GIST), uterine cancer, or ovarian cancer. In some aspects, the cancer is a hematologic cancer. In some aspects, the cancer is myeloma, multiple myeloma, B-cell lymphoma, follicular lymphoma, lymphocytic leukemia, or myelogenous leukemia.

In treating a subject having or suspected of having cancer as described herein, one or more cancer treatments may be administered to the subject. The cancer treatment may be any suitable cancer treatment. One or more cancer treatments described herein can be administered to a subject at any suitable frequency (e.g., one or more times over a period ranging from days to weeks). Examples of cancer treatments include, but are not limited to, surgical intervention, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy (e.g., chimeric antigen receptor and/or T cells with wild type or modified T cell receptor), targeted therapies such as administration of kinase inhibitors (e.g., kinase inhibitors that target specific genetic lesions such as translocations or mutations) (e.g., kinase inhibitors, antibodies, bispecific antibodies), signal transduction inhibitors, bispecific antibodies or antibody fragments (e.g., biTE), monoclonal antibodies, immune checkpoint inhibitors, surgery (e.g., surgical excision), or any combination of the above. In some aspects, cancer treatment may reduce the severity of cancer, alleviate symptoms of cancer, and/or reduce the number of cancer cells present in a subject.

In some aspects, the cancer treatment may be a chemotherapeutic agent. non-limiting examples of chemotherapeutic agents include: amsacrine (amsacrine), azacytidine (azacitidine), azathioprine (axathioprine), bevacizumab (or an antigen binding fragment thereof), bleomycin (bleomycin), busulfan (busulfan), carboplatin (carboplatin), capecitabine (capecitabine), chlorambucil (chlorambucil), cisplatin (cispratin), cyclophosphamide (cyclophosphamide), Cytarabine, dacarbazine (dacarbazine), daunorubicin (daunorubicin), docetaxel (docetaxel), doxifluridine (doxifluridine), doxorubicin (doxorubicin), epirubicin (epirubicin), erlotinib hydrochloride (erlotinib hydrochloride), etoposide (etoposide), fedarabine (fiudarabine), fluorouridine (floxuridine), Fludarabine (fludarabine), fluorouracil (fluorouracil), gemcitabine, hydroxyurea (hydroxyurea), idarubicin (idarubicin), ifosfamide (ifosfamide), irinotecan (irinotecan), lomustine (lomustine), nitrogen mustard (mechlorethamine), melphalan (melphalan), mercaptopurine (mercaptopurine), methotrexate (methotrxate), Mitomycin (mitomycin), mitoxantrone (mitoxantrone), oxaliplatin (oxaliplatin), paclitaxel (paclitaxel), pemetrexed (pemetrexed), procarbazine (procarbazine), all-trans retinoic acid, streptozotocin (streptozocin), tafluporin (tafluposide), temozolomide, teniposide (teniposide), thioguanine (tioguanine), topotecan (topotecan), uramestin (uramustine), valrubicin (valrubicin), vinblastine (vinblastine), vincristine (vincristine), vindesine (vindesine), vinorelbine (vinorelbine), and combinations thereof. additional examples of anti-cancer therapies are known in the art; see, for example, the American Society of Clinical Oncology (ASCO), the european society of oncology (ESMO), or the national integrated cancer network (NCCN) guidelines for therapy.

In monitoring a subject having or suspected of having cancer as described herein, monitoring can be performed before, during, and/or after the course of a cancer treatment. The monitoring methods provided herein can be used to determine the efficacy of one or more cancer treatments and/or select subjects for enhanced monitoring.

In some aspects, monitoring may comprise conventional techniques capable of monitoring one or more cancer treatments (e.g., efficacy of one or more cancer treatments). In some aspects, a diagnostic test (e.g., any of the diagnostic tests disclosed herein) may be administered to a subject selected for enhanced monitoring at an increased frequency compared to a subject not selected for enhanced monitoring. For example, diagnostic tests may be administered to subjects selected for enhanced monitoring twice daily, once daily, twice weekly, once weekly, twice monthly, once quarterly, once every half year, once annually, or any frequency therein.

In various aspects, the DNA is present in a biological sample taken from a subject and used in the methods of the invention. The biological sample may be virtually any type of biological sample that contains DNA. The biological sample is typically a fluid, such as whole blood or a portion thereof having circulating cfDNA. In embodiments, the sample comprises DNA from a tumor or a liquid biopsy such as, but not limited to, amniotic fluid, aqueous humor, vitreous humor, blood, whole blood, fractionated blood, plasma, serum, breast milk, cerebrospinal fluid (CSF), cerumen (cerumen), chyle, chyme, endolymph, perilymph, stool, breath, gastric acid, gastric juice, lymph, mucus (including nasal drainage and mucus), pericardial fluid, peritoneal fluid, pleural fluid, pus, nasal discharge, saliva, expired air condensate, sebum, semen, sputum, sweat, synovial fluid, tears, vomit, prostatic fluid, nipple aspirate fluid, tears, sweat, cheek swab, cell lysate, gastrointestinal fluid, biopsy tissue, and urine or other biological fluid. In one aspect, the sample comprises DNA from circulating tumor cells.

As disclosed above, the biological sample may be a blood sample. The blood sample may be obtained using methods known in the art, such as finger pricks or phlebotomys. Suitably, the blood sample is about 0.1ml to 20ml, or alternatively about 1ml to 15ml, wherein the volume of blood is about 10ml. Lesser amounts may also be used, as well as circulating free DNA in the blood. Microsampling and sampling by needle penetration biopsy, catheter, excretion or production of body fluids containing DNA is also a potential biological sample source.

The methods and systems of the present disclosure utilize nucleic acid sequence information, and thus may include any method or sequencing device for performing nucleic acid sequencing, including nucleic acid amplification, polymerase Chain Reaction (PCR), nanopore sequencing, 454 sequencing, insert tag sequencing. In some aspects, the methods or systems of the present disclosure utilize systems such as those provided by: as a result of Meana (Illumina, inc) (including but not limited to HiSeq^TMX10、HiSeq^TM 1000、HiSeq^TM 2000、HiSeq^TM 2500、Genome Analyzers^TM、MiSeq^TM、NextSeq、NovaSeq 6000 systems), applied biosystems Life technologies (Applied Biosystems Life Technologies) (SOLiD ^TM systems, ion PGM ^TM sequencer, ion Proton ^TM sequencer) or Genapsys (Genapsys) or BGI MGI (BGI MGI). Nucleic acid analysis may also be performed by a system provided by: oxford nanopore technology company (Oxford Nanopore Technologies) (GridiON ^TM、MiniON^TM) or pacific bioscience company (Pacific Biosciences) (Pacbio ^TM RS II or sequence I or II).

The present invention encompasses systems for performing the steps of the disclosed methods and is described in part in terms of functional components and various processing steps. Such functional components and processing steps may be realized by any number of components, operations, and techniques configured to perform the specified functions and achieve the various results. For example, the invention may employ various biological samples, biomarkers, elements, materials, computers, data sources, storage systems and media, information gathering techniques and processes, data processing standards, statistical analysis, regression analysis, and the like, which may perform various functions.

Accordingly, the present invention further provides a system for monitoring, detecting, analyzing and/or assessing cancer. In various aspects, the system comprises: (a) A sequencer configured to generate a low coverage whole genome sequencing dataset of a sample; and (b) a computer system and/or processor having functionality for performing the methods of the present invention.

In some aspects, the computer system further comprises one or more additional modules. For example, the system may include one or more of an extraction unit and/or a separation unit operable to select an appropriate gene component analysis (e.g., cfDNA fragments of a particular size).

In some aspects, the computer system further includes a visual display device. The visual display device may be operable to display a curve fit line, a reference curve fit line, and/or a comparison of the two.

The methods for detection and analysis according to the various aspects of the invention may be implemented in any suitable manner (e.g., using a computer program operating on a computer system). As discussed herein, an exemplary system according to aspects of the invention may be implemented in conjunction with a computer system, such as a conventional computer system including a processor and random access memory, e.g., a remotely accessible application server, web server, personal computer, or workstation. The computer system also suitably includes additional memory devices or information storage systems, such as mass storage systems and user interfaces, e.g., conventional monitors, keyboards, and tracking devices. However, the computer system may comprise any suitable computer system and associated devices, and may be configured in any suitable manner. In one embodiment, the computer system comprises a stand-alone system. In another embodiment, the computer system is part of a computer network that includes a server and a database.

The software needed for receiving, processing and analyzing information may be implemented in a single device or in multiple devices. The software is accessible via a network such that the storing and processing of information occurs remotely with respect to the user. The system according to various aspects of the present invention, as well as its various elements, provide functions and operations that facilitate detection and/or analysis, such as data collection, processing, analysis, reporting, and/or diagnostics. For example, in this aspect, a computer system executes a computer program that can receive, store, search, analyze, and report information related to the human genome or region thereof. The computer program may include a plurality of modules that perform various functions or operations, such as a processing module for processing raw data and generating supplemental data and an analysis module for analyzing the raw data and the supplemental data to generate a quantitative assessment of disease state models and/or diagnostic information.

The programs executed by the system may include any suitable process to facilitate analysis and/or cancer diagnosis. In one embodiment, the system is configured to build a disease state model and/or determine a disease state of the patient. Determining or identifying a disease state may include generating any useful information about the patient's condition relative to the disease, such as performing a diagnosis, providing information that aids in the diagnosis, assessing the stage or progression of the disease, identifying a condition that may indicate susceptibility to the disease, identifying whether further testing, predicting and/or assessing the efficacy of one or more therapeutic procedures may be recommended, or otherwise assessing the patient's disease state, likelihood of disease, or other health aspect.

The following examples are provided to further illustrate the advantages and features of the present invention, but are not intended to limit the scope of the invention. While this example is typical of the examples that may be used, other procedures, methods, or techniques known to those skilled in the art may alternatively be used.

Example 1

Detecting cancer and predicting total survival

In this example, the methods of the present disclosure are used to detect cancer and predict total patient survival. Appendix A sets forth the study and results.

This study of prospective enrolled individuals demonstrated the ability of cfDNA fragmentation assays to distinguish patients with cancer from individuals not with cancer. The assay of the present invention may exhibit high performance in a multi-cancer environment using only fragmentation-related information obtained from low coverage WGS.

The results indicate that machine learning models can use cfDNA fragmentation patterns to distinguish between cancer and non-cancer, despite the presence of common non-malignant conditions (including cardiovascular disease, autoimmune disease, or inflammatory disease). In addition, individuals with a higher DELFI score had poorer prognosis, independent of other characteristics.

These data support the development of whole genome cfDNA fragmentation assays for non-invasive detection of both single and multiple cancers.

Example 2

Free DNA fragmentation profile analysis for monitoring therapeutic response in metastatic cancer

The fraction of circulating tumor DNA (ctDNA) molecules in plasma (ctDNA burden) has become a viable measure for describing the total tumor burden of patients with cancer. ctDNA burden can vary over time, decreasing in response to treatment and increasing in tumor resistance to therapy. Monitoring ctDNA dynamics throughout the treatment may enable the surgeon to make treatment decisions in a timely manner. Ideally, there is a need for a rapid, low cost and universally applicable monitoring test that predicts treatment success and patient prognosis. Plasma ctDNA from liquid biopsies has great potential as a micro-invasive biomarker for tumor detection and response monitoring to (targeted) therapies. Plasma ctDNA is a dynamic tumor marker due to its short half-life and recurrence can be detected earlier than imaging and clinical parameters.

There are a variety of techniques for ctDNA profiling. Targeting Next Generation Sequencing (NGS) is a sensitive method that can provide information about somatic abnormalities and detect genomic changes in tumors. However, this approach has limitations due to the prevalence of cloned hematopoietic variants within the aging population. Tissue or leukocyte-guided methods must be used to prevent these variants from obscuring detection of tumor-specific changes. On the other hand, single nucleotide variants can be longitudinally tracked using less expensive ctDNA hot spot mutation methods such as drop digital PCR (ddPCR). Because these methods detect a limited number of somatic tumor changes, these hot spot mutation assays are generally not suitable for a wide range of tumors within a patient population and provide a narrow view of the genetic make-up of the tumor. For example, in patients with metastatic colorectal cancer (mCRC), trackable RAS/BRAF driving mutations are present in only half of patients.

Previous studies have shown that the size distribution of free DNA (cfDNA) in the genome can be exploited to reveal its origin. The proportion of shorter fragments is greater in people with cancer than in healthy people. Since the survival of patients with cancer is inversely related to the stage of disease, cfDNA fragment size composition (i.e., the ratio of shorter cfDNA fragments to longer cfDNA fragments) is used to develop tools for early disease detection in patients with cancer. This method, called DELFI (DNA evaluation of early intercept fragments), can distinguish cancer from non-cancer and indicates the origin of the tumor. Due to the minimally invasive nature of the technology, cfDNA fragment histology can also have additional clinical value for monitoring disease progression. Thus, DELFI tumor score (DELFI-TF), a machine-learned classifier that is capable of detecting tumor dynamics without the need for genetic information about the tumor of origin, was developed. In this work, the monitoring of therapeutic response of DELFI-TF classifier in patients with mCRC was evaluated.

DELFI-TF model development using whole genome cfDNA fragmentation profile-692 consecutive plasma samples from patients with mCRC and RAS/BRAF mutant (n=79) or RAS/BRAF wild type (n=74) disease that participated in prospective phase III clinical trials were processed and analyzed (CAIRO 5) (table 1, fig. 9A and fig. 9B). After study qualification (including the unresectable state of liver metastasis defined by the central panel), the mutant status of KRAS (exons 2,3 and 4), NRAS (exons 2 and 3) and BRAF (codon 600) were assessed in available tissue samples, followed by blood drawing at pre-treatment baseline and at successive time points during treatment (fig. 9A and 10). During the initial two month period, patients in the mutant and wild type subgroups were treated with a fluoropyrimidine-based first line regimen (FOLFOX or FOLFIRI) plus bevacizumab (fig. 9A and 10A). Thus, radiological evaluations intended to assess resectability of liver metastases were performed by central tumor committee review. Tumor-aware cfDNA analysis was performed retrospectively using droplet digital PCR (ddPCR) in 312 samples from patients in the mutant panel. Tumor-agnostic DELFI-TF analysis was successfully performed in 692 samples from patients in both the mutant and wild-type panels (fig. 9B and 10A). The DELFI-TF failure rates associated with library preparation and Whole Genome Sequencing (WGS) were 0.42% and 0.29%, respectively (FIG. 9B). Participants were scheduled for follow-up until death or study exit.

Table 1: study of demographic and clinical Properties of participants

OS, total lifetime; PFS, progression-free survival

To evaluate mutation independence for cancer specific changes in cfDNA, a DELFI-TF model was first designed (fig. 10B). For all cfDNA samples of patients in the mutant panel (n=312), tumor burden was initially quantified as Mutant Allele Frequencies (MAF) of RAS/BRAF variants as evidenced by tumor tissue measured by ddPCR. Full genome fragment sequencing statistics were obtained by low coverage WGS of cfDNA library using replicate aliquots of cfDNA samples (fig. 10B). The bayesian hierarchical regression model was trained and cross-validated for MAF of tumor-specific driven RAS/BRAF variants measured by ddPCR in all longitudinal cfDNA samples sequenced in the mutant panel. To generate predicted DELFI-TF values for each sample, this model considered DELFI score, plasma Aneuploidy (PA) score, and weight components from the mixed model that utilized cfDNA fragment size density (fig. 10B). Unsupervised cluster analysis was performed using a short-to-long ratio of segment sizes across 504 5-Mb bins and arm-level copy number z scores across 39 chromosome arms at baseline and at the time point in treatment for 128 patients with mCRC (fig. 10C). Notably, fragmentation pattern differences can be observed across several clinical and demographic properties at baseline time points in multiple regions throughout the genome of most patients with mCRC, which most correspond to high DELFI-TF values. Similarly, it was observed that most time points associated with progressive disease confirmed by imaging assessment also exhibited significant heterogeneity and high tumor scores, in contrast to most time points associated with stable disease or radioactive response after initiation of first-line systemic therapy, which was associated with fewer genomic abnormalities and DELFI-TF values (fig. 10C and 10D). These findings indicate that models designed to predict tumor scores in cfDNA can identify systemic treatment responses in real time in a non-invasive manner.

DELFI-TF accurately reflects cfDNA mutant allele frequencies and copy number changes-independent analysis was performed on DELFI-TF model using non-cancer control samples (n=155) from the group of symptomatic patients of denmark that were previously negatively checked for cancer diagnosis. Non-cancer control samples exhibited significantly lower DELFI-TF values with a 95% Confidence Interval (CI) upper limit of 0.006 compared to untreated samples (n=128) from the CAIRO group. Notably, DELFI-TF values were significantly higher than 0.006 from untreated samples from patients with mCRC (fig. 11A). Similar distributions were observed when samples from both groups were evaluated with ichorCNA, a tool designed to estimate tumor scores in ultra low pass WGS data (fig. 12A). However, the non-cancer control showed a broader range of tumor score values, including a higher upper 95% ci limit (0.017). In addition, ichorCNA values for a few CRC samples overlapped with the estimated tumor score for the non-cancer control, indicating that DELFI-TF more accurately reflects the disease burden detected in cfDNA samples.

Next, the analytical performance of DELFI-TF compared to the mutation-based tumor burden assessment was quantitatively evaluated for RAS/BRAFMAF using ddPCR (FIG. 11B). A strong correlation was observed between MAF values and DELFI-TF values (pearson, r=0.85, p=3.92 e-89). Interestingly, the range of positive DELFI-TF values observed in 60 (20.5%) treatment time points of undetectable RAS/BRAFMAF obtained by ddPCR was broad, meaning that the DELFI-TF method may be more sensitive to measuring tumor fraction in patients receiving effective anti-tumor treatment. It was further confirmed that these non-uniform time points where RAS/BRAFMAF was undetectable exhibited DELFI-TF values correlated with tumor scores predicted by ichorCNA (spearman, rho=0.54, p < 0.001) (fig. 12B). The correlation between cfDNA fragment groups and copy number changes in tissue (fig. 11C) samples and plasma samples (fig. 11D) was examined. Patient-matched formalin-fixed paraffin-embedded (FFPE) tumor tissue of 104 of 153 patients was analyzed by low-pass WGS (mean coverage 0.2×). It was consistently observed that abnormal cfDNA fragmentation patterns were present in copy neutral regions in the genome, and that regions of copy number variation in tissue samples were further affected (fig. 11C). In addition, it was also shown that there was an equal analytical performance and a significant correlation between tumor scores assessed by DELFI-TF and ddPCR MAF, with respect to cfDNA copy numbers of MBD1 (pearson, DELFI-TF: r= 0.64,p<0.001;ddPCR MAF:r = -0.67, p < 0.001) and PLGC1 (pearson, DELFI-TF: r= -0.84, p <0.001; ddpcrmaf: r=0.55, p < 0.001), genes were typically deleted and amplified in mCRC, respectively (fig. 11D). The ability to detect nucleosome depleted regions using relative coverage at the Transcription Start Site (TSS) as surrogate markers for gene expression in CRC was explored. It was observed that for the samples at baseline, the relative coverage at TSS of the group consisting of approximately 900 genes highly expressed in CRC was significantly lower than for the samples under treatment (wilcoxon, p < 0.001) (fig. 11E and 11F), reflecting that dynamic changes were detected during disease response and progression (fig. 12C). In summary, DELFI-TF accurately captured cancer-specific changes associated with MAF and copy number aberrations in cfDNA.

DELFI-TF was compared to clinical features and standard imaging assessment correlation-DELFI-TF method to clinical patient characteristics. At the untreated time point, a moderate correlation was observed between DELFI-TF and ddPCR MAF and the sum of the longest diameters of target metastatic lesions in the liver (SLD) (DELFI-TF, spearman, rho= 0.49,p<0.001;ddPCR MAF, spearman rho=0.48, p < 0.001) (fig. 13A). In contrast, no correlation was seen with the serum carcinoembryonic antigen (CEA) levels measured prior to initiation of treatment (DELFI-TF, spearman, rho=0.1, p=0.43; ddpcr MAF, spearman, ho=0.15, p=0.24) (fig. 13B). Baseline DELFI-TF and ddPCRMAF tumor scores did correlate with clinical response because by two consecutive scans, pre-treatment levels were significantly lower in patients with a later confirmed partial or complete response than in patients with stable or progressive disease (wilcoxon, DELFI-TF, p <0.05; ddpcrmaf, p < 0.05) (fig. 13C). In addition, baseline tumor scores were significantly lower for patients considered resectable following systemic induction therapy (Kruskal-Wallis, DELFI-TF, p <0.05;ddPCR MAF,p<0.05) (fig. 13D), as was patients with heterogeneous disease (wilcoxon, p < 0.05) (fig. 13E). Notably, DELFI-TF and ddPCRMAF tumor scores were not different due to tumor flanking (fig. 14A) or RAS/BRAF mutation status (fig. 14B). At baseline, DELFI-TF and ddPCRMAF tumor scores were significantly lower in patients (non-progressors) who had never developed disease progression at any point during the CAIRO trial than in patients (once progressors) who developed disease progression at some point during treatment (fig. 14C). On the other hand, analysis using SLD of target liver lesions at baseline failed to distinguish between non-progressors and once progressors (fig. 14C). Furthermore, DELFI-TF proved to accurately track longitudinal disease load dynamics even at advanced time points in patients treated with treatment intended hepatectomy (fig. 13C). Overall, it was demonstrated that once progressors more commonly exhibited an increase in DELFI-TF values at early time points and disease progression at late time points compared to non-progressors (fig. 13G).

After verifying the analytical equivalence between DELFI-TF model and ddPCR assay for RAS/BRAF MAF evaluation, a decision was made to explore further the correlation of dynamic changes of DELFI-TF with clinical outcome. To accommodate the longitudinal evolution of the consecutive DELFI-TF values in a single score, a DELFI-TF slope was calculated, which DELFI-TF slope was defined as the slope of the line fitted to the DELFI-TF value using linear regression starting at the first blood biopsy time point after the start of treatment and ending at the time of disease progression confirmed by RECIST 1.1. Then a trend of lower DELFI-TF slope (fischer accurate test (Fisher exact test), p=0.1) was observed for patients experiencing partial or complete responses as their best overall response (fig. 14E). Overall, time analysis of DELFI-TF and ddPCR MAF showed comparable tumor dynamics (fig. 15A). For patients in the wild-type panel, only DELFI-TF values could be used for time analysis (FIG. 15B). Patients with DELFI-TF slope below median had a higher objective radioactivity response rate and a longer follow-up duration to first line treatment than patients with DELFI-TF slope above median (fig. 16).

The baseline DELFI-TF and DELFI-TF slopes were then correlated with survival outcomes. At baseline, patients with DELFI-TF values below the first quartile showed a longer median progression-free survival (PFS) than patients with DELFI-TF values above the first quartile (13.4 months versus 10.2 months, risk ratio [ HR ] = 1.77, 95% ci 1.12 to 2.78, log rank p=0.013) (fig. 17A). For the RAS/BRAF mutant panel, baseline MAF for tumor score evaluation showed similar differences in median PFS (14.4 months versus 8.3 months, hr=2.56, 95% ci 1.36 to 4.83, log rank p= 0.00272) (fig. 18A). Serum CEA levels at baseline were not predictive of disease progression or death (fig. 18B). Within the trial, patients were evaluated by an expert panel two months after initiation of therapy to assess resectability of liver metastases. This first clinical response assessment did not distinguish the survival differences between patients with partial responses and patients with stable disease (11.3 months versus 11.2 months, hr=1.13, 95% ci 0.79 to 1.61, log rank p=0.52) (fig. 18C). Patients with DELFI-TF slope below median exhibited longer PFS in the entire study population (13.4 months versus 10.4 months, hr=2.03, 95% ci 1.247 to 3.318, log rank p=3.76 e-3) (fig. 17B) and in patients experiencing persistent clinical benefit defined as objective response or stable disease longer than 12 months (16.7 months versus 13.3 months, hr=2.235, 95% ci 1.097 to 4.553, log rank p=.023) (fig. 17C). Patients with a DELFI-TF slope below the median also exhibited significantly longer overall survival (OS; 59.4 months versus 29.1 months, hr=3.05, 95% ci 1.58-5.90, log rank p= 5.135 e-4) (fig. 17D). It was also observed that survival could be further stratified by resected status, where OS results for patients with complete resection of primary tumor and liver metastases are superior to those of patients with incomplete or no resection (fig. 17D).

Liquid biopsy cfDNA analysis is a new promising clinical tool in cancer research. DELFI-TF score was developed, a fragment histology method that is able to quantitatively measure tumor burden and shows its potential for longitudinal disease monitoring in patients with mCRC.

Currently, liquid biopsy ctDNA assays for the presence of cancer rely primarily on detecting one or more somatic tumor changes. Different research advantages utilized cfDNA fragment histology features as surrogate features. Size selection of cfDNA molecules (i.e., selecting shorter cfDNA fragments over longer cfDNA fragments) in vitro and in silico can enrich ctDNA and enhance identification of genetic alterations in ctDNA. In addition, whole genome fragmentation patterns can facilitate tumor detection and identification of the tumor of origin. The novelty of cfDNA fragment histology approach is that low coverage whole genome sequencing of trace cfDNA can be used without the need to detect driver mutations to monitor the patient's response longitudinally.

Despite advances in diagnosis and treatment, most patients with mCRC relapse, which provides a biomarker for guiding the course of treatment for clinical needs. However, currently available follow-up methods (such as clinical imaging and serum CEA) have limited accuracy in detecting the viability of tumor tissue and assessing the effectiveness of the treatment shortly after the initiation of therapy, and are therefore challenging. Current studies indicate that DELFI-TF may be more sensitive than conventional methods for therapy response monitoring, because DELFI-TF is more predictive of PFS than serum CEA measurements and clinical Computed Tomography (CT) imaging after treatment begins. Identifying the response or progression of the treatment provides the surgeon with the opportunity to adjust the patient's treatment regimen.

In addition to DELFI-TF, ddPCR MAF after initiation of treatment also predicts disease recurrence. However, the ability to detect differences in PFS in patients where ddPCR MAF is undetectable suggests that DELFI-TF may be more sensitive to therapy response monitoring, but the methods of fragment-histology monitoring may not track therapy-induced genomic changes in tumors, which is possible with targeted sequencing methods. In addition, both ddPCR MAF and DELFI-TF before treatment indicated the success rate of complete excision of liver metastases and OS. However, DELFI-TF has a conceptual advantage over hot spot mutation assays (e.g., ddPCR). DELFI-TF is generally suitable for use in samples of patients with any type of cancer, since it does not require prior knowledge of the driving changes of the tumor. The low coverage WGS required for fragmentation profile is lower cost than targeted sequencing. Since tumor burden can fluctuate over time, decreasing in response to treatment and increasing with tumor development resistance to therapy, DELFI-TF can be used as a tool for highlighting the appropriate time for more detailed sequencing analysis of the kit.

In a limited number of patients with post-hepatectomy blood samples, postoperative positives DELFI-TF appear to indicate disease recurrence and moderate sensitivity. However, pre-surgical blood testing may be a confounding factor in the training cohort. Measurements within 48 hours after surgery showed a spike in DELFI-TF. Since surgery is an invasive procedure, samples taken too close to the time of surgery may be indicative of wound healing, not tumour derived cfDNA. Therefore, the cut-off value of positive DELFI-TF results in samples extracted after complete excision (i.e., minimal residual disease environment) should be further studied.

Here, DELFI-TF was evaluated and orthogonal validation was applied at the sample level to a single nucleotide variant genotyping method derived from patients with mCRC using samples collected in well-controlled clinical trials. Thus DELFI-TF was defined and showed in the training set its potential prognostic power over conventional methods for monitoring of therapeutic response to detect disease progression. Note that these results must be confirmed in a validated group and that other types of cancer or early stage disease must also be evaluated later prior to clinical use. These results cannot be transferred directly into other body fluids such as urine and cerebrospinal fluid because of the different distribution of cfDNA fragments. In summary, novel quantitative measures of ctDNA burden were developed using cfDNA fragment histology. DELFI-TF appears to be a non-invasive method useful for monitoring the success of treatment of patients with mCRC within a training cohort.

Study design and population: the present study was a retrospective analysis of liquid biopsies collected from a homogeneous group of patients with mCRC who participated in a prospective CAIRO clinical trial (NCT 02162563). Phase III randomized CAIRO trial study the best first line systemic therapy for patients with histologically demonstrated CRC and isolated previously untreated initial unresectable liver metastases. Patients treated with dual chemotherapy (FOLFOX or FOLFIRI) and bevacizumab and at least one blood draw before and after treatment were included in the study. All patients were considered unresectable when included, i.e., R0 excision was not possible in one procedure with one surgical intervention. Following treatment with dual chemotherapy and bevacizumab, an expert panel consisting of liver surgeons and abdominal radiologists every two months followed current clinical practice to evaluate the likelihood of local treatment of colorectal liver metastases in patients. Clinical follow-up was performed according to the standard of care, which included clinical reviews every three months and CT imaging and serum CEA every six months. While liver metastasis remained unresectable, chemotherapy was continued without targeting agent, the total duration of pre-and post-operative treatment was six months, and patients were continuously assessed by serum CEA and CT imaging every two months until disease progression. Follow-up was recorded to day 1 of 2021, 9. The trial was approved by the medical ethics committee, performed according to the declaration of helsinki (Declaration ofHelsinki), and patients signed written informed consent for study participation and blood collection for conversion studies.

Blood collection and cfDNA extraction-collection of liquid biopsy samples was performed every three months prior to study treatment (baseline), pre-operative, post-operative, and during follow-up until disease progression or treatment was completed. Using 10mL of freeBlood samples were withdrawn from tubes (Style, inc. of Lavinita, U.S.A. (Streck, la Vista, USA)) and collected centrally at the Netherlands Institute of cancer (NETHERLANDS CANCER Institute) (Amsterdam, netherlands). The two-step centrifugation process (1700 Xg 10 min and 20000 Xg 10 min) separated the free plasma. The free plasma was stored at-80 ℃ until further use. cfDNA isolation was performed using QIAsymphony (Qiagen, germany) at an elution volume of 60 μl. cfDNA concentrations were assessed using QubitTM dsDNA high sensitivity assay (sameifeier, thermoFisher; waltherm, MA, USA) in MA. As input to library preparation, up to 15ng aliquots were prepared and added to 51 μl using TE buffer as necessary. cfDNA aliquots were shipped to the laboratory of Delfi diagnostic company (Delfi Diagnostics) (Baltimore, MD, USA) in maryland, USA.

Library preparation and cfDNA sequencing-after arrival at the laboratory, the extracted cfDNA was quantified using TapeStation 4200 (agilent technologies (Agilent Technologies); santa clara (SANTA CLARA, CA, USA) of california, usa). Preparation of NEBNEext DNA library kit (New England Biolabs (NEW ENGLAND Biolabs); isplasivelqi (Ipswich, mass., USA)) an NGS library was constructed with up to 15ng cfDNA input as described previously (19) with four major modifications to the manufacturer's guidelines: 1) Library purification step using AMPure XP on beads (Beckman Coulter); braia (break, CA, USA)) method in california to minimize sample loss during elution and tube transfer steps; 2) NEBNEext end repair, dA tailing and adaptor ligase and buffer volumes were adjusted as appropriate to accommodate the on-bead AMPure XP strategy; 3) Use of the factor Mener double index adapter in ligation reactions; and 4) using Phusion HotStart polymerase (Sieimer's Feier company; woltherm, massachusetts, U.S.A.) amplified cfDNA libraries for four cycles. The WGS library quality was determined using a 2100 bioanalyzer (agilent technologies; santa clara, california, usa) or TapeStation 4200 (agilent technologies; santa clara, california, usa). Next, a total of 96 double index cfDNA libraries containing samples with different barcodes were pooled together into a single lane of an S4 flow cell and 100-bp paired end (200 cycles) WGS sequencing was performed on NovaSeq 6000 (San Diego, CA, USA; san Diego, CA) with the goal of 8 x coverage per genome. To limit the batch effect, libraries of all samples taken from the same individual are created in the same batch, containing duplicate libraries as controls between batches and technical replicates of nucleosome DNA obtained from nuclease digested human peripheral blood mononuclear cells as controls within the batch.

RAS/BRAF mutation analysis-RAS and BRAF V600E mutation analysis was performed on tumor tissue DNA following routine clinical practice for all patients. For a subset of patients with RAS/BRAF tumor tissue mutations, longitudinal liquid biopsy hot spot mutation analysis and fragmentation analysis were performed by ddPCR (Bio-Rad, hercules, CA, USA). Screening kits were used according to manufacturer's instructions using ddPCR^TM KRAS G12/G13(#1863506)、ddPCR^TM KRAS Q61(#12001626)、ddPCR^TM KRAS A146T(#10049550) and ddPCR ^TM BRAF V600 (# 12001037), using 9. Mu.L of sample, 11. Mu.L of ddPCR probe method super mix (without dUTP), 1. Mu.L of multiplex assay and 1. Mu.L of nuclease free water. All measurements were performed in duplicate, including a blank (nuclease-free water) and a positive control. Patients with RAS/BRAF mutations that could not be tracked by ddPCR were excluded (fig. 9B). The data was analyzed using QuantaSoft ^TM software version 1.6.6 (bure corporation of heracles, california) and the auto-correction algorithm described previously.

CfDNA sequencing data analysis-paired-end sequenced reads were aligned to the reference genome (hg 19) on a per sample basis using paired-end alignment with Bowtie (version 2.3.0). Aligned reads were ordered and converted to BAM format using Samtools (version 1.3.1) and Bedtools (version 2.26.0) and then to BED format. The fragment length was calculated based on the start and end coordinates and the fragment was divided into 504 5-Mb bins covering approximately 2.6Gb of the genome. Next, the number of short (100-150 bp) and long (151-220 bp) fragments per bin was calculated using R/Bioconductor (version 3.6.2) and these counts were corrected by GC content as described by Benjamini and Speed. The corrected counts of short segments are divided by the corrected counts of long segments per bin to obtain a fragmentation pattern for each person.

Four statistics were calculated for each sample to generate DELFI-TF scores, DELFI scores, DELFI divergences, mixed model components, and arm level aneuploidy scores. The DELFI score was calculated similarly to the method described by Cristiano et al (Genome-wide cell-free DNA fragmentation IN PATIENTS WITH CANCER) for patients with cancer, nature, 570, 385-389 (2019), and the score indicated how similar the fragmentation pattern looked to individuals with or without cancer. DELFI divergence is defined as the correlation of a binned, average centered short-to-length ratio minus a given sample to a binned, average centered short-to-length ratio of a "median healthy" sample from a reference group containing only healthy samples. The hybrid model summarizes the segment size distribution and the weight statistics from this model are evaluated when DELFI-TF is generated.

Using these calculated statistics for each sample, a bayesian hierarchical regression model was trained using R for allele frequencies of tumor-specific driver RAS/BRAF variants measured by ddPCR in longitudinal cfDNA samples. This model takes DELFI scores, DELFI divergences, mixed model weights, and aneuploidy scores as inputs and outputs the predicted MAF. The model assumes that MAF is a random variable of β distribution and that the expected MAF for a given sample is functionally related to the described features by the inverse of the feature matrix multiplied by the vector of regression coefficients plus a patient-specific random intercept that accounts for intra-patient correlation between measurements. To generate bias-free predictions, overfitting was avoided and universality was assessed, training was done by leaving one patient out of cross-validation. In such a cross-validation scheme, data for each patient is set aside in turn, the model is trained on the remaining samples, and the trained model is then used to generate predictions for the set aside samples. DELFI-TF is defined as the predicted MAF generated by the cross-validation scheme. The quality of the predictions generated is assessed by evaluating the correlation of these predictions with observed ddPCR MAF values and evaluating the relationship between these predictions and time until progression or death.

DELFI-TF dynamic analysis-to capture molecular dynamics of tumor burden over time, DELFI-TF slope, i.e. slope of regression line fitted to DELFI-TF values within 60 days after the date of progression of OS analysis until before progression of PFS analysis after time T1, was calculated. For this practice, patients with at least 3 samples collected prior to progression were selected, and at least one of those samples was collected in a progression window of 120 days ago until the date of progression of PFS analysis (79 patients) and 120 days ago until 60 days after progression of OS analysis (80 patients). Regression lines were calculated using Python/scikit-learn (3.9.13/version 1.1.1).

Calculation of relative coverage for gene expression analysis—for this analysis, a collection of 854 transcripts identified by the Broad GDAC Firehose pipeline were selected, which are known to be highly expressed in colon adenocarcinoma, and whose Transcription Start Site (TSS) coordinates were extracted. Fragment coverage was calculated for these TSSs plus flanking regions of 1,500 bp on each side of all genes in only 126 patients with plasma samples at both time points T0 and T1. The TSS coordinates and the list of aligned fragments are in BED format and coverage calculations are performed using pybedtools (0.9.0 version), i.e., the python interface of Bedtools.

Statistical analysis-correlation between DELFI-TF and ddPCR MAF was calculated using pearson correlation coefficients. Similarly, the correlation between DELFI-TF/ddPCR MAF and copy number ratio was calculated using the pearson correlation. Stuffman correlation test was used between DELFI-TF/ddPCR MAF and SLD and serum CEA. Except for survival analysis, all double sample hypothesis tests used the wilcoxon rank sum test. Tumor scores based on DELFI-TF and MAF were compared between resected states using the Kruskal-Wallis test. Lifetime analysis was performed using Mantel-Cox log rank test. Analysis was performed using R statistics software (version 4.2.1, statistical computing foundation of vienna australia (Foundation for Statistical Computing)). Unless otherwise indicated, the test is assumed to be double sided, with type 1 error of 5% for determining statistical significance.

Although the invention has been described with reference to the presently preferred embodiments, it should be understood that various modifications can be made without departing from the spirit of the invention. Accordingly, the invention is limited only by the following claims.

Claims

1. A method of monitoring cancer, the method comprising:

calculating a fragmentation score based on the cfDNA fragmentation profile, the score indicating the likelihood of cancer being present in the subject;

calculating a divergence score based on the ratio of the short segments to long segments in the sample as related to the ratio of samples from healthy subjects;

Determining a set of model weights based on the segment size distribution;

Determining, by the machine learning model, a monitoring score for the sample based on the fragmentation score, the divergence score, and the model weight, the monitoring score indicating a level of tumorigenic nucleic acids in cfDNA of the sample.

2. The method of claim 1, wherein the divergence score indicates a correlation between the fragmentation profile and a median fragmentation profile observed for healthy subjects.

3. The method of claim 1, wherein the monitoring score is in the range of 0 to 1.

4. The method of claim 1, further comprising determining at least one of a likelihood of total survival, a likelihood of progression-free survival, or a time until progression of the subject based on the monitoring score.

5. The method of claim 4, wherein the likelihood of the overall survival of the subject, the likelihood of progression-free survival, or the time to progression decreases as a monitoring score value increases.

6. The method of claim 5, further comprising classifying the monitoring score as a high score or a low score, wherein a high score indicates at least one of a decrease in total survival, a decrease in progression-free survival, or a decrease in time until progression of the subject.

7. The method of claim 6, wherein the monitoring score is categorized prior to administration of a cancer treatment to the subject to provide a baseline categorization of the subject.

8. The method of claim 6, wherein the monitoring score is categorized at a first time point after administration of a cancer treatment to the subject to provide a post-treatment categorization of the subject.

9. The method of claim 1, wherein the likelihood of cancer progression in the subject increases as the monitoring score value increases.

10. The method of claim 1, wherein the likelihood of cancer progression in the subject decreases as the monitoring score value decreases.

11. The method of claim 1, wherein the likelihood of a positive response to a cancer treatment administered to the subject increases as the monitoring score decreases.

12. The method of claim 1, wherein the likelihood of a positive response to a cancer treatment administered to the subject decreases as the monitoring score increases.

13. The method of claim 1, wherein the monitoring score predicts a score of a tumor-derived nucleic acid or a clonal mutant allele fraction of a cancer-associated mutation.

14. The method of claim 1, wherein the cancer is a solid tumor.

15. The method of claim 1, wherein the cancer is a sarcoma, carcinoma, or lymphoma.

16. The method of claim 1, wherein the cancer is selected from the group consisting of: colorectal cancer, lung cancer, kidney cancer, brain cancer, prostate cancer, breast cancer, pancreatic cancer, bile duct cancer, liver cancer, CNS cancer, gastric cancer, esophageal cancer, gastrointestinal stromal tumor (GIST), uterine cancer and ovarian cancer.

17. The method of claim 1, wherein the cancer is a hematological cancer.

18. The method of claim 1, wherein the cancer is selected from the group consisting of: myeloma, multiple myeloma, B-cell lymphoma, follicular lymphoma, lymphocytic leukemia, and myelogenous leukemia.

19. The method of claim 1, further comprising administering a cancer treatment to the subject.

20. The method of claim 17, wherein the cancer treatment is selected from the group consisting of: surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormone therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, or any combination thereof.

21. The method of claim 1, wherein the cfDNA fragmentation profile is determined by:

obtaining cfDNA fragments from the subject and isolating the cfDNA fragments;

Sequencing the cfDNA fragments to obtain sequenced fragments;

mapping the sequenced fragments to a genome to obtain a window of mapped sequences; and

Analyzing a window of the mapped sequences to determine cfDNA fragment lengths and generating the cfDNA fragmentation profile.

22. A method of determining at least one of total lifetime, progression-free lifetime, or time until progression, the method comprising:

Determining a ratio of short fragments to long fragments and fragment size distribution as related to the ratio of samples from healthy subjects from the fragmentation profile in the samples;

Determining a set of model weights based on the segment size distribution;

Determining, by the machine learning model, a monitoring score for the sample based on the fragmentation score, the divergence score, and the model weight, the monitoring score indicating a level of tumor origin in cfDNA of the sample, thereby indicating a likelihood of cancer progression in the subject; and

At least one of a total lifetime, a progression-free lifetime, or a time until progression is determined based on the monitoring score.

23. A system for monitoring cancer in a subject, the system comprising:

A memory; and

Determining a set of model weights based on the segment size distribution;

Determining, by the machine learning model, a monitoring score for the sample based on the fragmentation score, the divergence score, and the model weight, the monitoring score indicating a level of tumor origin in cfDNA of the sample, wherein the level of tumor origin in the cfDNA indicates a likelihood of cancer progression in the subject.